Last updated on August 19th, 2025 at 02:53 pm

In factories across the country, machines quietly log data.

How long a process takes. How much energy is used. The quantity of scrap or waste. The temperature inside the facility.

But without the ability to extract and connect this information, the numbers become meaningless.

“At least for small to medium manufacturers, this is a huge problem where there exists a lot of data and companies know enough to know they should be tracking and collecting this, but it’s often housed in a very cumbersome way,” said Reagan Cook, a PIN Fellow who recently applied her degrees in mechanical engineering and data analytics to help solve that problem at Atlanta-based Carbice.

Cook’s work was supported by Georgia Artificial Intelligence in Manufacturing (Georgia AIM), a statewide initiative that funds innovation and workforce development to strengthen the use of AI in manufacturing. Through the AI in Manufacturing Fellowship offered by the Partnership for Innovation (PIN) — a project in Georgia Tech’s Enterprise Innovation Institute — Cook spent six months exploring data challenges firsthand.

She first worked at the Georgia Tech Research Institute, helping develop a database of CAD drawings using AI. Then she joined Carbice, a company that manufactures heat-dissipating materials for electronics, where she was tasked with improving how the company collected and used its production data.

“In this case, the data is being collected off machines and the software the machine runs is proprietary, so we don’t have control over what the data looks like or how it’s collected,” Cook said. “And anything reported in-house is reported manually, so not many things were automated and some things aren’t formatted in a certain way.”

A Resource for Manufacturers

The PIN Fellowship is designed to solve challenges like this. By placing technical experts within private industry, it bridges the gap between academic research and real-world application. For companies like Carbice, the fellowship provides much-needed bandwidth without stretching internal staff.

“We have a process. I wouldn’t say it’s a bespoke process, but scaling is our main engineering challenge,” said Craig Green, chief technology officer at Carbice. “So that’s been the challenge—going from a lab at Georgia Tech to ATDC, to where we’re at now. We follow the same recipe but do it all over again, starting from zero.

“In doing that, we have generated a tremendous amount of data, both about the process and in inputs and outputs. And sometimes it’s difficult to use that data because it lives in spreadsheets—and it’s challenging to talk about the data if you’re not the one who is living with it all.”

This is where Cook’s fellowship made an impact. She focused on extracting key data from the company’s proprietary systems and organizing it into a central database that could be analyzed in the context of Carbice’s production process.

“The primary project was to capture the existing data into a place where it could be viewed outside of the proprietary software,” said Green. “From there, Reagan worked on pulling it and other data into a place where it can be viewed and analyzed within the context of the manufacturing process.”

The challenge isn’t unique to Carbice. Many manufacturers invest in data-logging machines long before they develop a coherent data strategy. As technologies evolve, so do the data collection methods. This creates a fragmented and often outdated approach to analytics.

Cook said having a dedicated fellow focused solely on solving this issue made a big difference.

“This is a small company, so people are pulled in different directions,” she said. “So I think it’s helpful for them where this is the only project I’m working on. I’m not tied down by other responsibilities, so it’s nice to offload this project on someone.”

For Carbice, the project unlocked new potential.

“Once we got over the initial step of creating and aggregating the data, it allowed us to follow up on other projects,” said Green. “So, we’ve been working on driving that roadmap and using some of the insights we’ve been able to gain by looking at a broad cross-section of data to determine what parameters we’ll optimize for our next series of tests. It’s been very helpful.”